Developing Tools for Species-Level Identification: Computer Vision Enables Rapid Identification of Wild Bee Species Using Wing Venation

Abstract

Through their role as pollinators, bees continue to be major contributors to the survival and productivity of both native plant communities and the agricultural industry. However, in spite of their ecological and economic importance, estimates of global bee populations have demonstrated rapid declines in species richness and density over the past century. Although these estimates provide information on general population trends, many of them lack specificity and offer little insight into dynamics occurring at the species-level. Additionally, the process of determining species-level identifications for bees relies heavily upon a shrinking number of expert taxonomists who are capable of recognizing subtle morphological differences between species. For these reasons, monitoring programs and large-scale conservation for insect pollinators has been slow, costly, and relatively ineffective. This study proposes the creation of a publicly available database of wing-images for the most common bee species found across New Jersey, US, and the creation of an application which will use these images to deduce accurate species-level identifications of insect pollinators. With the implementation of deep learning frameworks, an application for automatic identification of bee pollinators can be made available for use by researchers, students, and local communities. By generating a publicly available database of forewing images, future users will be able to expand the training set of our model to include new species, and ultimately increase the repertoire of species for which it can produce predictions. Through the integration of recent advances in machine learning technology and citizen science approaches to data collection, this project has the potential to both streamline the pollinator identification process and create opportunities for community outreach and global participation in efforts to conserve remaining populations of bees.